期刊名称：International Journal of Innovative Research in Computer and Communication Engineering
印刷版ISSN：2320-9798
电子版ISSN：2320-9801
出版年度：2016
卷号：4
期号：2
页码：1209
DOI：10.15680/IJIRCCE.2016.0402020
出版社：S&S Publications
摘要：This paper describe about a novel feature based on Local Shape Context descriptor for the obstacle detection on Railway tracks, this shape context descriptor is widely used algorithm in the object recognition, but it does not match for the object complex situations. Why because it does not consider the edges orientations. To address this issue, this present work add the edge based orientation information to the shape context descriptor. For that, this work first compute the image gradient in the nine directions and it can beextractthe shape context descriptor in each direction. Finally it can be put the feature vector to linear SVM for training. Then it test this descriptor's performance on the real-time video captured by the thermal night vision camera. The experimental results show that this work achieved a high detection rate and had fewer dimension than the other descriptors
关键词：Thermal Night Vision; Local Shape Context Descriptor; SVM
